Glossary

LLM - Large language model, such as gpt or Llama families. They work on text, and usually text only.

Stable Diffusion - This is typically your image generation models, or anything visual

Custom LLM - A LLM that you have customised in some way. Usually in one of the two following ways. a) Fine tuning b) Prompt engineering

Fine-tuning - Modifying the weights of an existing LLM, with the goal of adding data or efficiency, usually in 1 specific topic

Prompt-Engineering - This is a quicker version to get specific results of ( fine-tuning, but separate in technical execution), by adding specific context and instructions before a question, prompt-engineering allows to get more consistent answers, without the need of fine-tuning

Character - This is the term we use for the templates we use to prompt, as they are set up to allow users to turn LLMs into "characters" or "job specialists"

Agents - An LLM(s) that are connected together with external tools, capable of completing tasks that are more extensive that simple chatting or data fetching.

Personalised Agents - Agents that have had specific data to help them get better in a task, usually by prompt-engineering, or by fine-tuning an LLM with custom data sets.

Prompting - Adding additional context to your messages to guide a model

Model - Synonym for either LLM or Stable Diffusion systems. I should not use it as such, but that is how I use it